Название: Fundamentals of Reinforcement LearningАвтор: Rafael Ris-Ala
Издательство: Springer
Год: 2023
Страниц: 103
Язык: английский
Формат: pdf (true), epub
Размер: 18.8 MB
Artificial Intelligence (AI) applications bring agility and modernity to our lives, and the reinforcement learning technique is at the forefront of this technology. It can outperform human competitors in strategy games, creative compositing, and autonomous movement. Moreover, it is just starting to transform our civilization.
This book provides an introduction to AI, specifies Machine Learning techniques, and explores various aspects of reinforcement learning, approaching the latest concepts in a didactic and illustrated manner. It is aimed at students who want to be part of technological advances and professors engaged in the development of innovative applications, helping with academic and industrial challenges.
Understanding the Fundamentals of Reinforcement Learning will allow you to:
Understand essential AI concepts
Gain professional experience
Interpret sequential decision problems and solve them with reinforcement learning
Learn how the Q-Learning algorithm works
Practice with commented Python code
Find advantageous directions
Chapter 1 introduces the research area of Artificial Intelligence, as well as distinguishing between the various Machine Learning approaches and the types of problems they solve. The meaning of Reinforcement Learning is playfully introduced with examples, and its framework is explained. Then, relevant historical milestones that permeate several sciences and that have contributed to the development of this line of research are addressed.
Chapter 2 covers the fundamental knowledge needed to understand the entire system that involves Reinforcement Learning. Concepts such as agent, environment, actions, rewards, policies, and value function are discussed. Examples and analogies are presented to help illustrate each of these concepts, from structuring problems starting from the Markov Chain through Watkins and Dayan’s proposal and unfolding in the Bellman Equation. Finally, the classes and particularities of algorithms that have been successful in this innovative field of research are presented.
Chapter 3 illustrates the step-by-step operation of one of the most widely used algorithms in Reinforcement Learning, the Q-Learning algorithm. The meaning of each component of the algorithm and its demonstration through pseudocode is presented. Then, a detailed explanation of how the algorithm works is described through a visual example of an agent interacting in an environment, from the initialization of the Q-Table to the agent’s decision-making based on its experiences with the environment, through the construction of a policy to be followed.
Chapter 4 deals with practical tools for developing solutions in Reinforcement Learning. Some main libraries and frameworks are available for implementing RL algorithms, such as TensorFlow, Keras, and OpenAI Gym. Some useful data sources for conducting your RL experiments are also discussed.
In Chap. 5, a practical case of developing an autonomous cab with AI in Python is proposed. The details of the environment are discussed, and the action of the agent without the use of AI is exemplified. As a counterpoint, next, how to implement an RL algorithm is demonstrated in a simplified way. The code is commented on and explained in detail, illustrating the differences and advantages of using RL in these types of problems. The system is made available for further testing and implementation.
Chapter 6 presents the most recent applications of how Reinforcement Learning is impacting various areas of knowledge. Examples of RL applications in areas such as robotics, games, education, and quantum mechanics are presented. The main advantages and challenges of RL applications in different fields are also discussed, as well as perspectives for the future of RL use in each of these areas.
Скачать Fundamentals of Reinforcement Learning